基于Res-CNN和燃油压力波的柴油机喷油器故障诊断方法  

A Method for Fault Diagnosis of Fuel Injector of Diesel Engine Based on Res‐CNN and Fuel Pressure Wave

作　　者：靳莹 乔新勇 顾程 郭浩 宁初明 Jin Ying;Qiao Xinyong;Gu Cheng;Guo Hao;Ning Chuming(Department of Vehicle Engineering,Army Academy of Armored Forces,Beijing 100072;Unit 66407 of Chinese PLA,Beijing 100089;Institute of Military Engineering and Technology,Institute of System Engineering,Academy of Military Science,Beijing 100010)

机构地区：[1]陆军装甲兵学院车辆工程系,北京100072 [2]中国人民解放军66407部队,北京100089 [3]军事科学院系统工程研究院军需工程技术研究所,北京100010

出　　处：《汽车工程》2021年第6期943-951,共9页Automotive Engineering

基　　金：武器装备维修改革项目(2015WX05)资助。

摘　　要：燃油喷射系统的工作质量直接影响柴油机工作过程及性能。针对利用燃油压力波进行故障诊断时压力波特征点自动化识别困难、影响实时在线监测的问题,提出了利用深度学习图像识别理论进行喷油器故障诊断的方法。通过喷油泵试验台进行了喷油器典型故障模拟试验,测取了高压油管燃油压力波,分析了不同故障状态下燃油压力波动特征及规律,建立了基于深度残差的卷积神经网络(Res-CNN)模型,以一维燃油压力波信号为输入,进行喷油器故障诊断检测及验证,并对故障特征学习过程进行了可视化分析。结果表明,该模型较传统方法具有更高的诊断准确率,验证了直接应用燃油压力波图形识别方法进行在线实时监测的可行性。The working quality of the fuel injection system influences the working process and performance of diesel engine directly.Due to the difficulty in recognizing the characteristic points in fuel pressure wave automatically by using fuel pressure wave for fault diagnosis,the on‐line fault diagnosis is influenced.This paper proposes a fuel injector fault diagnosis method by deep learning image recognition theory.Experiments are conducted on a fuel injection pump test bench to simulate typical faults,and the fuel pressure wave of high‐pressure fuel pipe is measured.The fuel pressure wave characteristics and laws under different fault conditions are analyzed.A deep residual CNN network(Res‐CNN)model is built to detect and verify the faults,with the one‐dimension fuel pressure wave signal as the input,and the learning process of fault characteristics is analyzed visually.The results show that the model has higher diagnostic accuracy than the traditional method,which verifies the feasibility of direct application of fuel pressure wave image recognition method for on‐line real‐time monitoring.

关 键 词：柴油机 喷油器 故障诊断 深度学习 卷积网络 

分 类 号：TK428[动力工程及工程热物理—动力机械及工程]